Growth Metrics & Unit Economics Questions

The core quantitative vocabulary of growth: activation, retention, referral and revenue metrics, growth-accounting frameworks such as AARRR, and unit economics including LTV, CAC, and payback period. Covers defining these metrics precisely and computing growth calculations that reveal whether growth is efficient and durable. The scope is metric definitions and economic math, not go-to-market execution.

HardTechnical

146 practiced

A BI dashboard shows LTV increasing but churn rate is creeping up in a specific region. As the BI analyst, how would you reconcile these conflicting signals and what additional metrics or decomposition would you present to stakeholders?

Sample Answer

Situation: The dashboard shows LTV rising while churn in one region is increasing — an apparent contradiction that could mislead stakeholders.

Approach / Task: I would reconcile these signals by decomposing LTV and churn by cohorts, segments, and revenue drivers, verify data accuracy, and present a clear set of metrics and visualizations that explain how both can be true.

Actions I would take and what I'd present:- Data quality checks: confirm definitions (churn window, LTV formula), verify regional mapping and time alignment (revenue lag vs churn measurement).- Cohort LTV vs cohort churn: show LTV by acquisition month cohort alongside their monthly retention curves — rising LTV overall can be driven by older cohorts while recent cohorts exhibit higher churn.- Revenue decomposition: break LTV into ARPU (avg revenue per user), purchase frequency, average order value, and customer lifetime. Show which component is growing (e.g., higher spend by power users) despite rising attrition among casual users.- Segment analysis: churn by persona, product SKU, channel, price tier, and device. Highlight if high-value customers are sticking while low-value volumes drop (or vice versa).- Time-lagged analysis: compare current churn (recent months) to LTV calculated over longer windows; annotate the dashboard to show that LTV is backward-looking.- Leading indicators: present product usage, engagement metrics (DAU/MAU, session length), support tickets, NPS/CSAT, promo/discount exposure — to explain cause of churn spike.- Sensitivity/what-if: show how a sustained churn increase by X% in the region would impact projected LTV and revenue.

Result / Recommendation: I’d deliver an executive dashboard with cohort retention charts, component decomposition, and recommended actions (targeted retention campaigns for affected segments, pricing/UX experiments, monitoring cadence). This clarifies that rising LTV can coexist with localized churn and provides actionable next steps to mitigate risk.

MediumTechnical

99 practiced

Decompose subscription revenue into a formula using acquisition, conversion, and retention submetrics (express revenue as combination of these terms). For a SaaS product, explain how a 1% change in retention rate versus a 1% change in acquisition would affect quarterly revenue—show numeric example with baseline numbers.

HardSystem Design

81 practiced

Design a BI system to compute daily rolling 12-month LTV per acquisition channel for a company with 50M customers and 2B transactions. Describe data model (tables/partitions), incremental computation strategy, caching/aggregation layers, and how you'd keep dashboards performant in Looker/Tableau/Power BI.

Sample Answer

Requirements & constraints:- Daily rolling 12-month LTV per acquisition channel for 50M customers and 2B transactions.- Low-latency dashboards (interactive filters), accurate daily refresh, cost-efficient incremental updates.

High-level architecture:- Raw ingestion: streaming (Kafka) + batch (daily S3/Cloud Storage) -> raw_events table in data lake.- ETL/processing: Spark/Dataproc or Snowpark jobs to transform into analytical tables.- Serving: Columnar warehouse (BigQuery / Snowflake / Redshift Spectrum) + aggregated OLAP store (Druid/Pinot or materialized tables).

Data model:1. raw_transactions (date, txn_id, customer_id, amount, channel, attrs) — partitioned by date, compressed Parquet/ORC.2. customers (customer_id, acq_date, acq_channel, cohort_attrs) — clustered by acq_date.3. customer_daily_agg (date, customer_id, channel, revenue, orders) — partitioned by date and bucketed by customer_id for small file distribution.4. ltv_rollup_daily (as_of_date, acq_month, acq_channel, window_start, window_end, customers_count, total_revenue, avg_ltv) — partitioned by as_of_date and acq_channel.

Incremental computation strategy:- Use CDC / daily change detection: process only new/changed transactions and new customers.- Compute customer_daily_agg incrementally by upserting daily partitions (append + dedupe by txn_id).- For rolling 12m LTV: maintain a sliding-window materialized view per (acq_month, as_of_date, channel) computed by summing customer_daily_agg over last 365 days. Implement using incremental rolling-sum: - Store per-customer cumulative revenue by acq_month and date. - On each daily run, add revenue for new day and subtract revenue 366 days ago (if exists), updating ltv_rollup_daily by group-by — avoids full recompute.- Use idempotent jobs and watermarking to handle late-arriving data; backfill micro-batches as needed.

Aggregation / caching layers:- Pre-aggregate at multiple granularities: - Daily per-acq_channel & acq_month (for default dashboard granularity). - Weekly/monthly aggregates for longer-range queries.- Expose aggregates as materialized tables in warehouse with clustering on acq_channel and as_of_date for fast predicates.- Cache hot queries in a query cache (BigQuery BI Engine / Snowflake result cache / Druid segments).- Serve real-time needs via OLAP engine (Druid) for sub-second group-bys over recent periods.

Dashboard performance strategies:- Limit live row scans: dashboards query pre-aggregated ltv_rollup_daily rather than raw_transactions.- Use parameterized queries and persistent derived tables (PDTs) in Looker or extract refreshes in Tableau/Power BI.- Implement federated extracts for large user loads: scheduled daily extracts for executive dashboards; on-demand caches for analysts.- Use query time filters: require channel or time-bucket selection to avoid full-scan.- Materialize common filter combinations (top channels, last 90/365 days).- Optimize visuals: avoid heavy table calculations in BI tool; push calculations to warehouse SQL; use pagination and sampled previews.- Monitor & autoscale: track slow queries, add clustered indexes, increase compute for windowed jobs during nightly windows.

Trade-offs & resilience:- Incremental rolling-sum reduces compute but requires storing day-offsets per customer (storage vs CPU trade-off).- OLAP segment engines cost more but give interactive latency.- Plan for late-arriving transactions with micro-batch reprocessing and lineage for auditability.

Key metrics to monitor: job latency, staleness, query 95th-percentile latency, cost per TB scanned, data freshness SLA.

HardTechnical

79 practiced

Explain how regulatory constraints (GDPR, CCPA) and cookie deprecation affect LTV by channel measurement and attribution. Propose mitigation strategies a BI team can implement to preserve reliable unit-economics reporting.

Sample Answer

Regulatory constraints (GDPR/CCPA) and browser cookie deprecation reduce availability of deterministic user-level identifiers and third‑party signals. That breaks per-channel attribution and inflates LTV uncertainty because: (1) cross‑device/session stitching fails, (2) ad platforms can’t report full conversion paths, (3) sampling/attribution windows change — all leading to biased channel ROI and misestimated unit economics.

Mitigation strategies for a BI team:

1. Shift measurement from deterministic to aggregated & modeled: - Report cohort LTVs (acquisition-week or acquisition-campaign cohorts) rather than per-user LTV. - Use probabilistic attribution models (multi-touch attribution with Bayesian or Markov models) and expose uncertainty bands.

2. Use first‑party data & server‑side instrumentation: - Implement server-side event collection and robust event schemas (consent-aware). - Capture first‑party identifiers (logged-in user IDs, hashed emails with consent) and UTM parameters at acquisition.

3. Privacy-safe integrations: - Use clean rooms (Ads Data Hub, Snowflake clean rooms) for join-on-hashed identifiers to get aggregated attribution without raw PII leaving partners. - Leverage platform-provided privacy-safe measurement (e.g., SKAdNetwork, privacy‑preserving attribution APIs).

4. Causal/experimental methods: - Run uplift/incrementality tests and holdout experiments to validate modeled attribution and directly measure channel marginal ROAS.

5. Governance & instrumentation: - Tag a canonical acquisition funnel, enforce tracking QA, version event schemas, and track consent rates to adjust models. - Surface confidence intervals and “assumptions” metadata in dashboards so stakeholders understand limits.

6. Monitoring & reconciliation: - Reconcile modeled LTVs with finance (revenue recognition) monthly; monitor drift and recalibrate.

Practical example: build weekly acquisition-cohort LTV dashboard (by source/UTM) showing 7/30/90-day revenue per cohort with modelled attribution shares and a separate column with experimental uplift where available. This preserves actionable unit-economics while complying with privacy and browser limits.

MediumTechnical

97 practiced

Compute per-unit gross margin for a product where: list price = $80, average discounts = 10%, variable fulfillment cost = $12, product COGS = $18, and allocated monthly fixed overhead per unit at current volume = $5. Show the per-unit gross margin and comment whether fixed overhead should be included in 'gross margin' for pricing decisions.

Unlock Full Question Bank

Get access to hundreds of Growth Metrics & Unit Economics interview questions and detailed answers.

Join thousands of developers preparing for their dream job.